feat: add working Codex node launch/teardown

src/codex.bash

@@ -5,11 +5,24 @@ LIB_SRC=${LIB_SRC:-$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)}
 source "${LIB_SRC}/config.bash"
 # shellcheck source=./src/utils.bash
 source "${LIB_SRC}/utils.bash"
+# shellcheck source=./src/procmon.bash
+source "${LIB_SRC}/procmon.bash"
 
 _cdx_output=$(clh_output_folder "codex")
 _cdx_logs="${_cdx_output}/logs"
 _cdx_data="${_cdx_output}/data"
-_cdx_binary="${CLH_CODEX_BINARY:-codex}"
+
+if [ -z "${CODEX_BINARY}" ]; then
+  run command -v codex
+  _cdx_binary="${output}"
+else
+  _cdx_binary="${CODEX_BINARY}"
+fi
+
+if [ ! -f "${_cdx_binary}" ]; then
+  echoerr "Error: no valid Codex binary found"
+  exit 1
+fi
 
 _cdx_base_api_port=8080
 _cdx_base_disc_port=8190
@@ -17,6 +30,8 @@ _cdx_base_metrics_port=8290
 
 _cdx_node_start_timeout=30
 
+declare -A _cdx_pids
+
 cdx_cmdline() {
   local api_port\
     disc_port\
@@ -58,7 +73,7 @@ cdx_cmdline() {
   # shellcheck disable=SC2140
   echo "${cdx_cmd}"\
 " --log-file=${_cdx_logs}/codex-${node_index}.log --data-dir=${_cdx_data}/codex-${node_index}"\
-" --api-port=${api_port} --disc-port=${disc_port} --loglevel=INFO"
+" --api-port=${api_port} --disc-port=${disc_port} --log-level=INFO"
 }
 
 cdx_get_spr() {
@@ -75,20 +90,42 @@ cdx_get_spr() {
 }
 
 cdx_launch_node() {
-  _check_codex_binary
+  _cdx_ensure_outputs 0 || return 1
 
   local codex_cmd
-  codex_cmd=$(cdx_cmdline "$@")
+  codex_cmd=$(cdx_cmdline "$@") || return 1
 
   (
     $codex_cmd
     pm_job_exit $?
   )&
   pm_track_last_job
+  _cdx_pids[$1]=$!
 
   cdx_ensure_ready "$@"
 }
 
+cdx_destroy_node() {
+  local node_index="$1" wipe_data="${2:-false}" pid
+  pid="${_cdx_pids[$node_index]}"
+  if [ -z "$pid" ]; then
+    echoerr "Error: no process ID for node $node_index"
+    return 1
+  fi
+
+  # Prevents the whole process group from dying.
+  pm_stop_tracking "$pid"
+  pm_kill_rec "$pid"
+  await "$pid" || return 1
+
+  unset "_cdx_pids[$node_index]"
+
+  if [ "$wipe_data" = true ]; then
+    rm -rf "${_cdx_data}/codex-${node_index}"
+    rm -rf "${_cdx_logs}/codex-${node_index}.log"
+  fi
+}
+
 cdx_ensure_ready() {
   local node_index="$1" timeout=${2:-$_cdx_node_start_timeout} start="${SECONDS}"
   while true; do
@@ -104,4 +141,10 @@ cdx_ensure_ready() {
 
     sleep 0.2
   done
+}
+
+_cdx_ensure_outputs() {
+  local node_index="$1"
+  mkdir -p "${_cdx_logs}" || return 1
+  mkdir -p "${_cdx_data}/codex-${node_index}" || return 1
 }

src/procmon.bash

@@ -35,6 +35,11 @@ pm_start() {
         fi
 
         exit_code=$(cat "${_pm_output}/${pid}.pid")
+        if [ $? -ne 0 ]; then
+          echoerr "[procmon] ${pid} died with unknown exit code. Aborting."
+          _pm_halt "halted_no_return"
+        fi
+
         if [ -z "$exit_code" ]; then
           echoerr "[procmon] ${pid} died with unknown exit code. Aborting."
           _pm_halt "halted_no_return"
@@ -165,8 +170,13 @@ pm_join() {
 }
 
 pm_job_exit() {
-  exit_code=$1
-  echo "$exit_code" > "${_pm_output}/${BASHPID}.pid"
+  local pid_file="${_pm_output}/${BASHPID}.pid" exit_code=$1
+  # If the process is not tracked, don't write down an exit code.
+  if [ ! -f "$pid_file" ]; then
+    echoerr "[procmon] no PID file found for process $BASHPID"
+  else
+    echo "$exit_code" > "${_pm_output}/${BASHPID}.pid"
+  fi
   exit "$exit_code"
 }
 

test/test_codex.bats

@@ -50,15 +50,26 @@ setup() {
   rm -rf "$data_dir"
 }
 
-# @test "should launch a Codex node" {
-#   export CLH_CODEX_BINARY="${BATS_TEST_DIRNAME}/codex/build/codex"
+@test "should launch a Codex node" {
+  pm_start
 
-#   assert cdx_launch_node 0
-#   assert cdx_ensure_ready 0 1
+  assert cdx_launch_node 0
+  assert cdx_ensure_ready 0 3
 
-#   # We should see a log file and a data directory.
-#   assert [ -f "${_cdx_output}/logs/codex-0.log" ]
-#   assert [ -d "${_cdx_output}/data/codex-0" ]
+  # We should see a log file and a data directory.
+  assert [ -f "${_cdx_output}/logs/codex-0.log" ]
+  assert [ -d "${_cdx_output}/data/codex-0" ]
 
-#   cdx_destroy_node 0
-# }
+  pid="${_cdx_pids[0]}"
+  assert [ -n "$pid" ]
+
+  cdx_destroy_node 0 true
+
+  refute [ -d "${_cdx_output}/data/codex-0" ]
+  refute [ -f "${_cdx_output}/logs/codex-0.log" ]
+  assert [ -z "${_cdx_pids[0]}" ]
+
+  assert $(! kill -0 "$pid")
+
+  pm_stop
+}

test/test_helper/common_setup.bash

@@ -4,10 +4,4 @@ common_setup() {
   load test_helper/bats-assert/load
 
   export LIB_SRC="${BATS_TEST_DIRNAME}/../src"
-  export CODEX_BINARY=${CODEX_BINARY:-"${BATS_TEST_DIRNAME}/codex/build/codex"}
-
-  if [ ! -f "$CODEX_BINARY" ]; then
-    echo "Codex binary not found at $CODEX_BINARY"
-    exit 1
-  fi
-}
+}